Every rejected Roman numeral form
Seven rule violations, and 1952 additive spellings between 1 and 3999 that are not the canonical numeral.
Free to download and reuse under CC BY 4.0. Credit Romanize, https://romanize.net.
Every converter will tell you what a valid Roman numeral looks like. This page publishes the other half: the forms that get rejected, why each one is rejected, and what it should have been. The converter here works that out on every conversion anyway, because it round-trips each result back to the number and refuses anything that does not return identical, so the list below is not an opinion about the rules. It is what the engine decides.
Download
Columns: invalid_form, value, canonical, rule, reason. The JSON carries the same content with the rule catalogue and the generated forms as separate arrays, plus the licence and attribution fields. Both are regenerated on every build from the same function that powers the converter.
The seven ways a numeral can be wrong
| Rejected | Should be | Rule broken |
|---|---|---|
| IIII | IV | A symbol repeats more than three times. I, X, C and M may appear at most three times in a row. A fourth repetition is written with the subtractive pair instead. |
| VV | X | V, L or D is repeated. The half-step symbols never repeat, because two of them always make the next symbol up. VV is X, LL is C, DD is M. |
| IL | XLIX | A symbol subtracts that is not allowed to. Only I, X and C subtract. V, L and D never do. IL, VC and LM look reasonable and are not numerals. |
| IC | XCIX | A subtraction skips more than one step. A subtracting symbol may only precede the next one or two larger symbols. I goes before V and X, X before L and C, C before D and M. Nothing else. |
| IXIX | XVIII | The same subtractive pair appears twice. A subtractive pair is a single unit and each one may appear at most once in a numeral, in its own place value. |
| IVX | not a number | Groups run small to large. After a subtractive pair the numeral must continue with a smaller place value. IVX is not a number at all: it parses as neither 14 nor 6. |
| MMMM | (IV) | The value falls outside 1 to 3999. The seven letters stop at 3999. Above that a vinculum (a bar over the numeral, commonly typed as brackets) multiplies it by a thousand. There is no numeral for zero or for a negative number. |
The additive spellings, 1 to 100
The largest category by far is the additive spelling: the numeral written without ever using a subtractive pair. It is not a modern invention. Roman inscriptions regularly show IIII and XXXX, and the strict six-pair system was only fixed much later. It is still the single commonest way a numeral comes out wrong today, because writing out the tally is the intuitive thing to do.
Below is the part of the set that falls between 1 and 100. The full 1952 rows to 3999 are in the download.
| Rejected | Value | Canonical |
|---|---|---|
| IIII | 4 | IV |
| VIIII | 9 | IX |
| XIIII | 14 | XIV |
| XVIIII | 19 | XIX |
| XXIIII | 24 | XXIV |
| XXVIIII | 29 | XXIX |
| XXXIIII | 34 | XXXIV |
| XXXVIIII | 39 | XXXIX |
| XXXX | 40 | XL |
| XXXXI | 41 | XLI |
| XXXXII | 42 | XLII |
| XXXXIII | 43 | XLIII |
| XXXXIIII | 44 | XLIV |
| XXXXV | 45 | XLV |
| XXXXVI | 46 | XLVI |
| XXXXVII | 47 | XLVII |
| XXXXVIII | 48 | XLVIII |
| XXXXVIIII | 49 | XLIX |
| LIIII | 54 | LIV |
| LVIIII | 59 | LIX |
| LXIIII | 64 | LXIV |
| LXVIIII | 69 | LXIX |
| LXXIIII | 74 | LXXIV |
| LXXVIIII | 79 | LXXIX |
| LXXXIIII | 84 | LXXXIV |
| LXXXVIIII | 89 | LXXXIX |
| LXXXX | 90 | XC |
| LXXXXI | 91 | XCI |
| LXXXXII | 92 | XCII |
| LXXXXIII | 93 | XCIII |
| LXXXXIIII | 94 | XCIV |
| LXXXXV | 95 | XCV |
| LXXXXVI | 96 | XCVI |
| LXXXXVII | 97 | XCVII |
| LXXXXVIII | 98 | XCVIII |
| LXXXXVIIII | 99 | XCIX |
The one exception worth knowing
IIII is rejected by the rules and is still correct on a clock face. That is not a contradiction, it is a convention that predates the rule and was never abandoned by clockmakers. The page on the Roman numeral clock sets out the competing explanations and is careful to mark which of them are documented and which are folklore. Where a claim is a convention rather than a hard fact, we say so.
Frequently asked questions
Is IIII a valid Roman numeral?
No. Four is IV. IIII is the older additive spelling, and it survives on clock faces by convention, but it fails the rule that no symbol repeats more than three times.
Why is IL not 49?
Because only I, X and C may be subtracted, and only from the next one or two larger symbols. I may precede V and X, never L. Forty-nine is XLIX.
How many invalid forms are there?
Infinitely many if you count every possible string of letters. The useful set is finite: seven categories of rule violation, plus the 1,952 additive spellings between 1 and 3999 that differ from the canonical numeral. Both are published here.
Can I reuse this dataset?
Yes. The CSV and JSON are free to use under CC BY 4.0. Credit Romanize, https://romanize.net.
Related
Last updated July 22, 2026. Reviewed for accuracy against the standard rules for Roman numerals. How we check accuracy.